CHAPTER 13 PROBABILISTIC RISK ANALYSIS The objectives are to: Use of statistical and probability concepts in decision making involving risk and uncertainty Illustrate how they can be applied in engineering economy analysis Discuss the limitations relative to their application RANDOM VARIABLES Factors having probabilistic outcomes The probability that a cost, revenue, useful life, or other economic factor value will occur, is usually considered to be the estimated likelihood that an event (value) occurs The information about these random variable that is particularly helpful in decision making are the expected values and variances These values for random variables are used to make the uncertainty associated with each alternative more explicit RANDOM VARIABLES Capital letters such as X, Y, and Z are used to represent random variables Lower-case letters (x,y,z) denote the particular values that these variables take on in the sample space (I.e., the set of possible outcomes for each variable When random variable X follows some discrete probability distribution, its mass function is usually indicated by p(x) and its cumulative distribution function by P(x) When X follows a continuous probability distribution, its probability density function and its cumulative distribution function are usually indicated by f(x) and F(x), respectively DISCRETE RANDOM VARIABLES A random variable X is discrete if it can take on a finite number of values (x1,x2…xL) The probability that a discrete random variable X takes on the value xi is given by Pr{X = xi} = p(xi) for i = 1,2,….,L (i is a sequential index of the discrete values, xi, that the variable takes on) where p(xi) > 0 and i p(xi) = 1 CONTINUOUS RANDOM VARIABLES A random variable is continuous if: c Pr{c < X < d} = f(x)dx d In the nonnegative function f(x),this is the probability that X is within the set of real numbers (c,d) f(x)dx = 1 - The probability that the value X is less than or equal x = k, the cumulative distribution function F(x) for a continuous case is d Pr{X < k} = F(k) = f(x)dx d - Pr{c < X < d} = f(x)dx = F(d) – F( c ) c In most applications, continuous random variables represent measured data, such as time, cost and revenue on a continuous scale MATHEMATICAL EXPECTATIONS AND SELECTED STATISTICAL MOMENTS The expected value of a single random variable X, (E(X), is a weighted average of the distributed values x that it takes on and is a measure of the central location of the distribution E(X) is the first moment of the random variable about the origin and is called the mean of the distribution E(X) = i xi p( xi ) for x discrete and i = 1,2,…,L E(X) = xf(x)dx for x continuous - MATHEMATICAL EXPECTATIONS AND SELECTED STATISTICAL MOMENTS From binomial expansion of [X – E(X)]2, it can be easily shown that: V(X) = E(X2) – [E(X)]2 V(X) is the second moment of the random variable around the origin : the expected value of X2, minus the square of its mean V(X) is the variance of the random variable X V(X) = i x2p(xi) – [E(X)]2 for x discrete V(X) = xi2f(x)dx – [E(X)]2 for x continuous The standard deviation of a random variable, SD(X) is the positive square root of the variance SD(X) = [V(X)]1/2 MULTIPLICATION OF A RANDOM VARIABLE BY A CONSTANT When a random variable, X, is multiplied by a constant, c, the expected value E(cX), and the variance, V(cX) are: E(cX) = cE(X) = cxi p(xi) for discrete E(cX) = cE(X) = cx f(x)dx for continuous V(cX) = E{ [cX – E(cX)]2 } =c2E{ [X – E(X)]2 } =c2 V (X) MULTIPLICATION OF TWO INDEPENDENT VARIABLES When a random variable, Z, is a product of two independent random variables, X and Y, the expected value, E(Z), and the variance, V(Z) are Z= XY E(Z) = E(X) E(Y) V(Z) = V(X) [E(Y)]2 + V(Y) [E(X)]2 + V(X) V(Y) EVALUATION OF PROJECTS WITH DISCRETE RANDOM VARIABLES Expected value and variance concepts apply theoretically to long-run conditions in which it is assumed that the event is going to occur repeatedly However, application of these concepts is often useful when investments are not going to be made repeatedly over the long run Problem 13-4 In a building project, the amount of concrete to be poured during the next week is uncertain. The foreman has estimated the following probabilities: Amount (cubic yards) 1000 1200 1300 1500 2000 probabilities 0.1 0.3 0.3 0.2 0.1 Determine the expected value (amount) of concrete to be poured next week. Also compute the variance and standard deviation of the amount of concrete to be poured. Solution: Let X = amount of concrete to be poured E(X) = 1000 (0.1) + 1200 (0.3) + … +2000 (0.1) = 1,350 cubic yards V(X) = [(1000)2 (0.1) + (1200)2(0.3) +…+(2000)2 (0.1)] – (1350)2 = 66500 SD(X) = (66500)1/2 = 258 EVALUATION OF PROJECTS WITH CONTINUOUS RANDOM VARIABLES Two Frequently Used Assumptions Uncertain cash-flow amounts are distributed according to the normal distribution Uncertain cash flow amounts are statistically independent No correlation between cash flow amounts is assumed EVALUATION OF PROJECTS WITH CONTINUOUS RANDOM VARIABLES If there is a linear combination of two or more independent cash flow amounts (i.e., PW = c0F0 + … +cNFN, where ck values are coefficients and Fk values are periodic net cash flows) the expression V(PW) reduces to N V(PW) = k=0 N E(PW) = ck2 V(Fk) k=0 ckE(Fk) Problem 13-6 A small dam is being planned for a river tributary that is subject to frequent flooding. From past experience, the probabilities that water flow will exceed the design capacity of the dam during a year, plus relevant cost information, are as follows: Design A B C D E Probability of greater flow during a year 0.10 0.05 0.025 0.015 0.006 Capital investment -$180,000 - 195,000 - 208,000 - 214,000 - 224,000 Estimated annual damages that occur if water flows exceed design capacity are $150,000, 160,000, 170,000, 190,000, and 210,000 for design A, B, C, D, and E, respectively. The life of the dam is expected to be 50 years, with negligible salvage value. For an interest rate of 8% per year, determine which design should be implemented. What nonmonetary considerations might be important to the selection? Solution: Prob. of Annual Total Greater Capital Estimated Expected Equivalent Flow during Investment Damage Cost of Present Design a year Damages Worth ----------------------------------------------------------------------------------------------------A 0.1 -180,000 -150,000 -15,000 -363,000 B 0.05 -195,000 -160,000 -8,000 -292,000 C 0.025 -208,000 -175,000 -4,375 -261,522 D 0.015 -214,000 -190,000 -2,850 -248,865 E 0.006 -224,000 -210,000 -1,260 -239,414 EVALUATION OF UNCERTAINTY USING MONTE CARLO SIMULATION Computer-assisted simulation tool for analyzing more complex project uncertainties Monte Carlo simulation generates random outcomes for probabilistic factors which imitate the randomness inherent in the original problem MONTE CARLO SIMULATION Can be used only when a process has a random, or chance, component Based on a probabilistic distribution Random samples taken from this probability distribution are analogous to observations made on the system itself As the number of observations increases, the results of the simulation will more closely approximate the random behavior of the real system BASIC STEPS IN MONTE CARLO SIMULATION Identify a probability distribution for each random component of the system Work out an assignment so that intervals of random numbers will correspond to the probability distribution Obtain the random numbers needed for the study Interpret the results SOURCE OF RANDOM NUMBERS Large studies Computer-generated random numbers Small studies Table of random digits Example The manager of a machine shop is concerned about machine breakdowns. A decision has been made to simulate to breakdowns for a 1-day period. Historical data on breakdowns over the last 100 days are given in the following table: Number of Breakdowns 0 1 2 3 4 5 frequency 10 30 25 20 10 5 Simulate breakdowns for a 10-day period. Solution Number of Breakdowns Frequency 0 10 1 30 2 25 3 20 4 10 5 5 Probability .10 .30 .25 .20 .10 .05 Cumulative Probability .10 .40 .65 .85 .95 1.00 Using the table of random digits 18 25 73 12 54 96 23 31 45 01 Corresponding Random numbers 00 to 09 10 to 39 40 to 64 65 to 84 85 to 94 95 to 99 Day Random Number Stimulated number of breakdowns -------------------------------------------------------------------------------------1 2 3 4 5 6 7 8 9 10 18 25 73 12 54 96 23 31 45 01 1 1 3 1 2 5 1 1 2 0 Mean number of breakdowns = (17/10) = 1.7 Expected number of breakdowns = 0*.1 + 1*.3+ 2*.25 +3*.2+ 4*.1+5*.05=2.05 per day SIMULATING THEORATICAL DISTRIBUTIONS Poisson distribution Exponential distribution Normal distribution Use a table of normally distributed random numbers See Table 13-10 Simulated value = mean + Random number x standard deviation Uniform distribution Simulated value = A + (RN/RNm )[B-A] where RNm = maximum possible random number RN = selected random number A = minimum outcome B = maximum outcome USING MONTE CARLO SIMULATION IN ENGINEERING PROJECTS Construct an analytical model that represents the actual decision situation Develop a probability distribution from subjective or historical data for each uncertain factor in the model Sample outcomes are randomly generated by using probability distribution for each uncertain quantity and then used to determine a trial outcome for the model Repeating sampling process many times leads to a frequency distribution of trial outcomes, which are used to make probabilistic statements EXAMPLE 13-9 The following estimates relate to an engineering project being considered by a large manufacturer of air-conditioning equipment. Subjective probability functions have been estimated for the four independent uncertain factors as follows: CAPITAL INVESTMENT Normally distributed with a mean of –$50,000 and a standard deviation of $1,000. USEFUL LIFE Uniformly and continuously distributed with a minimum life of 10 years and maximum life of 14 years ANNUAL REVENUE $35,000 with a probability of 0.4 $40,000 with a probability of 0.5 $45,000 with a probability of 0.1 ANNUAL EXPENSES Normally distributed, with a mean of -$30,000 and a standard deviation of $2000. The management of the company wishes to determine whether or not the capital investment in the project will be a profitable one. MARR = 10%. Solution Five trial outcomes are computed manually Random capital Project project Normal investment life, N life, N Trail deviate [-50,000+ [10+(RN/999) (nearest Number (RND1) RND1(10,000) 3-digit RNs (14-10)] integer) -----------------------------------------------------------------------------------------------------------1 +1.003 -$48,997 807 13.23 13 2 +0.358 - 49,642 657 12.63 13 3 -1.294 - 51,294 488 11.95 12 4 +0.019 - 49,981 282 11.13 11 5 -0.147 - 50,147 504 12.02 12 ------------------------------------------------------------------------------------------------------------ Annual revenue -$35,000 for 0-3 Annual Expense 1-digit -40,000 for 4-5 [-30,000 + PW RN 45,000 for 9 RND2 RND2(20,000) (R+E)(P/A,10,N) ----------------------------------------------------------------------------------------------------------1 2 $35,000 +0.036 -$29,928 -$12,969 2 0 35,000 +0.605 - 31,000 - 22,720 3 4 40,000 -1.470 - 32,940 - 3,189 4 9 45,000 -1.864 - 33,728 + 23,232 5 8 40,000 +1.223 - 27,554 + 34, 656 -----------------------------------------------------------------------------------------------------------Total +19,010 Estimated average present worth = (19,010/5) = $3,802 DECISION TREES Also called decision flow networks and decision diagrams Powerful means of depicting and facilitating analysis of important problems, especially those that involve sequential decisions and variable outcomes over time Practical tool because it permits large complicated problems to be reduced to a series of smaller simple problems Enable objective analysis and decision making that includes explicit consideration of the risk and effect of the future GENERAL PRINCIPLE OF DIAGRAMING The Decision Tree Diagram Should Show the Following (With square symbol to depict decision node and circle symbol to depict chance outcome node): All initial or immediate alternatives among which the decision maker wishes to choose All uncertain outcomes and future alternatives the decision maker wishes to consider Note alternatives at any point and outcomes at any chance outcome node must be: Mutually exclusive Collectively exhaustive; that is, one event must be chosen or something must occur if the decision point or outcome node is reached